Data subsetting
How to carve small, representative, referentially complete slices out of large databases, driven by the business entities your tests care about.
Topics
Subsetting and referential integrity
A subset is useful only if it is consistent: every order needs its customer, every claim its policy, across databases and files. These pages explain how entity-based subsetting keeps those relationships intact, how to size a subset before building it, and how a narrow subset helps investigate a production incident.
Start here
Go deeper
Referential integrity
Why every foreign key in a test dataset must resolve, and how masking and subsetting can quietly break joins across tables and systems.
Cross-system test-data subsetting
How to give QA one customer and every related policy, claim, account and transaction across several systems, without dangling references or unrelated records.
Subset-Size Calculator
Reproducing production issues with masked data
Recreate the records behind a production incident as a small, masked and certified dataset - seed lists stay in your network. A Preview that needs a newer agent release.
Relational subsetting without orphan rows
A runnable comparison of naive row sampling and an entity-driven subset on synthetic claims data, counting the dangling references each one leaves behind.
Data engineer